Scholarlink

<p>In modern cloud environments, efficient management of computational resources is a critical challenge due to the growing demand for scalable and high-performance applications. Horizontal scaling in Kubernetes (K8s) clusters is essential for dynamically adjusting resources to match workload demands. However, their reactive nature often limits traditional autoscaling methods like K8s Horizontal Pod Autoscaler (HPA), leading to inefficiencies under variable loads. To overcome these limitations, more advanced and adaptive scaling approaches are needed. Thus, this study introduces an adaptive approach to horizontal scaling in K8s clusters using Artificial Neural Networks (ANNs) for load forecasting, referred to as ANN-HS. The proposed method aims to enhance the efficiency of resource consumption and optimize replica allocation compared to the standard HPA. By leveraging pre-trained regression models, ANN-HS dynamically adjusts resources to meet varying demands, ensuring adherence to latency requirements and improving overall system performance. Experimental results demonstrate that ANN-HS outperforms traditional HPA methods, offering a scalable and flexible solution for managing microservices in cloud environments. This approach provides a robust framework for optimizing horizontal scaling in Kubernetes, contributing to the advancement of intelligent resource management in cluster computing. Experimental results show that ANN-HS significantly improves resource utilization compared to Kubernetes’ HPA. Specifically, ANN-HS reduces CPU consumption by approximately 50% while maintaining Service Level Agreement (SLA) compliance with an average violation rate of less than 10%. Additionally, ANN-HS reduces the number of replicas needed by 66.67%, optimizing resource allocation under varying load conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive horizontal scaling in kubernetes clusters with ANN-based load forecasting

  • 期刊论文

  • 发表日期：

    2025年1月21日

查看全文
  • Lucileide M. D. da Silva,
  • Pedro V. A. Alves,
  • Sérgio N. Silva,
  • Marcelo A. C. Fernandes

摘要

In modern cloud environments, efficient management of computational resources is a critical challenge due to the growing demand for scalable and high-performance applications. Horizontal scaling in Kubernetes (K8s) clusters is essential for dynamically adjusting resources to match workload demands. However, their reactive nature often limits traditional autoscaling methods like K8s Horizontal Pod Autoscaler (HPA), leading to inefficiencies under variable loads. To overcome these limitations, more advanced and adaptive scaling approaches are needed. Thus, this study introduces an adaptive approach to horizontal scaling in K8s clusters using Artificial Neural Networks (ANNs) for load forecasting, referred to as ANN-HS. The proposed method aims to enhance the efficiency of resource consumption and optimize replica allocation compared to the standard HPA. By leveraging pre-trained regression models, ANN-HS dynamically adjusts resources to meet varying demands, ensuring adherence to latency requirements and improving overall system performance. Experimental results demonstrate that ANN-HS outperforms traditional HPA methods, offering a scalable and flexible solution for managing microservices in cloud environments. This approach provides a robust framework for optimizing horizontal scaling in Kubernetes, contributing to the advancement of intelligent resource management in cluster computing. Experimental results show that ANN-HS significantly improves resource utilization compared to Kubernetes’ HPA. Specifically, ANN-HS reduces CPU consumption by approximately 50% while maintaining Service Level Agreement (SLA) compliance with an average violation rate of less than 10%. Additionally, ANN-HS reduces the number of replicas needed by 66.67%, optimizing resource allocation under varying load conditions.